Revista Ítalo-Española de Derecho Procesal
pp. 1-20
Madrid, 2026
DOI: 10.37417/rivitsproc/3317
Marcial Pons Ediciones Jurídicas y Sociales
© Esma Yağmur Sönmez
ISSN: 2605-5244
Recibido: 18/01/2026 | Aceptado: 18/05/2026
Editado bajo licencia Creative Commons Attribution 4.0 International License.
ALGORITHMIC NEUTRALITY OR DIGITAL INEQUALITY? AI AND THE LESSONS FROM ISDS
Dr. Esma Yağmur Sönmez *
Senior Expert
Turkish Radio Television Corporation
ABSTRACT: The article argues that AI-driven judicial reform initiatives within civil justice systems repeat an established pattern in which critique is absorbed into procedure. Using Investor–State Dispute Settlement (ISDS) as a cautionary mirror, it shows how managerial legality performs neutrality while reproducing inequality. It then outlines a human-centred alternative grounded in participation, accountability, and contestability, calling for a shift from design-led legitimacy to democratic oversight.
KEYWORDS: AI; Legitimacy; Accountability; ISDS; TWAIL
SUMMARY: 1. INTRODUCTION. LAW’S NEW TECHNICAL TURN.— 2. THEORETICAL FRAME. LEGITIMACY AND NEUTRALITY IN LAW.— 3. THE LESSONS OF ISDS. MANAGERIAL LEGALITY IN PRACTICE: 3.1. The Promise of Neutral Expertise; 3.2. The Reality of Asymmetry; 3.3. The Managerial Turn and Its Afterlife in AI.— 4. THE RISE OF DIGITAL JUSTICE. EFFICIENCY, ACCESS, AND ASYMMETRY: 4.1. Reform as Continuity: 4.1.1. Reform as Performance; 4.1.2. The Continuity of Managerial Legality; 4.1.3. The Discourse of Innovation; 4.1.4. Critical Re-reading. Reform as Continuity; 4.2. Legitimacy, Authority, and the Politics of Reform; 4.2.1. Law’s Double Claim. Authority and Legitimacy; 4.2.2. The Politics of Authority; 4.2.3. Algorithmic Legality and the Politics of Expertise; 4.3. Beyond Reform. Critical Perspectives.— 5. A HUMAN-CENTRED CIVIL JUSTICE: 5.1. From System Trust to Human Trust; 5.2. A Brief Analogy: Climate Justice and Digital Governance.— 6. CONCLUSION. FROM NEUTRALITY TO ACCOUNTABILITY.— BIBLIOGRAPHY.
Artificial intelligence (AI) has entered legal discourse with striking ease. Across jurisdictions, courts are experimenting with automated filing, digital case management, and predictive analytics. Governments present these systems as pathways to efficiency and access to justice, while private vendors market them as tools of modernisation. The vocabulary—transparency, objectivity, neutrality—echoes earlier moments when law sought to renew its legitimacy through technical improvement rather than political reform (Kennedy, 2016, pp. 93–100, 106–107; Koskenniemi, 2005, pp. 589–596).
Yet this technical turn also redefines what fairness means. When procedures become automated, judgment does not disappear—it moves into design. How datasets are constructed, optimisation goals defined, and acceptable margins of error determined all shape decisions that are not purely technical. In practice, these choices often reflect underlying political judgments (Crawford, 2021, pp. 134–136; Eubanks, 2018, pp. 43–47). What looks like neutrality is already an act of governance.
This article offers a conceptual comparison between the reform discourse of Investor–State Dispute Settlement (ISDS) and contemporary AI-driven judicial reform initiatives within civil justice systems. Both, it argues, perform legitimacy through the language of expertise and neutrality. The comparison is not intended to suggest institutional equivalence between ISDS and civil justice systems. Rather, it operates at the level of governance rationalities, focusing specifically on how AI-driven judicial reform initiatives translate legitimacy concerns into managerial and technical solutions. ISDS once presented arbitral professionalism as a guarantee of fairness; digital justice initiatives now present algorithmic precision in much the same way. In each case, reform transforms critique into management—what Koskenniemi (2005, pp. 589–596) describes as the technical management of political conflict.
Arbitration was often presented as a way to depoliticize economic relations. In practice, however, it has been associated with constraints on states’ regulatory autonomy and with reinforcing existing North–South asymmetries (Anghie, 2005, pp. 245–251; Miles, 2013, pp. 112–118; Van Harten, 2007, pp. 28–31). Reform projects that equate fairness with efficiency risk missing the point. Access and participation are not by-products of good design; they are its test. When they disappear, reform turns into performance. Authority is increasingly exercised through technical systems that structure and shape behaviour, often in ways that remain difficult to interrogate or contest (Yeung, 2018, pp. 507–510).
Digitalization in civil justice is taking shape within societies already marked by a deep digital divide. If these realities are ignored, technology will not close the gap—it will widen it in quieter ways. The critical traditions of international law help to name what is happening.
Koskenniemi reminds us that law often swings between moral hope and political compromise, and TWAIL asks whose history and experience are hidden when neutrality is taken for granted. Read together, they show that good intentions easily turn into management routines: reform begins to look like change, even when the structure stays the same.
Seen through this lens, the promise of human-centred AI reflects the same optimism that once surrounded ISDS reform—the belief that fairness can be engineered. The challenge, however, is not to perfect design but to recover accountability, to ensure that the authority law exercises through technology remains open to question.
This article proceeds as follows. Part II develops the theoretical frame drawing on Koskenniemi and TWAIL; Part III examines how managerial legality shaped ISDS and its reforms; Part IV analyses how similar logics appear in digital justice architectures; Part V outlines a human-centred alternative grounded in participation and rights; and the conclusion argues that neutrality must give way to accountability if legitimacy is to endure in the age of AI.
Legitimacy is never something law can simply rely on; it has to be rebuilt through rules, procedure, and the promise of fairness. Legal orders sustain authority by managing conflict through neutrality and expertise. This is the core of Martti Koskenniemi’s account of international law, where argument moves endlessly between utopia and apology—between moral hope and political compromise (Koskenniemi, 2005, pp. 589–596). When law is accused of bias, it retreats into technical form; when charged with rigidity, it borrows the language of justice. Each move repairs legitimacy for a moment, yet the tension remains. Law often manages conflict by translating it into the language of principle rather than resolving it outright.
In the digital age, this oscillation returns in technological form. AI tools in courts transform moral questions into design choices: accuracy rates, optimisation goals, explainability standards. The result is a restatement of Koskenniemi’s account of how legal reasoning manages political conflict through technical form. The promise of human-centred AI, much like the older faith in “rules-based adjudication,” performs impartiality while keeping disagreement manageable. What matters is not only the outcome but the image of method: a system that looks consistent is presumed legitimate even when its foundations remain opaque (Kennedy, 2016, pp. 93–100).
Third World Approaches to International Law (TWAIL) bring a second strand to this argument. TWAIL scholars remind us that neutrality has a history. What claims to be universal often carries the marks of hierarchy — for instance, colonial investment protections once framed as “development”. Emerging in response to colonial legal order, TWAIL shows how doctrines such as civilization or development masked domination behind legal reason (Anghie, 2005, pp. 245–251; Mutua, 2000, p. 32). In this light, neutrality is not an escape from politics but its continuation in another language. The same logic now risks reappearing in digital justice reform, where inequality is reframed as a technical flaw in data or design rather than as a question of power (Gathii, 2011, pp. 27–29).
Koskenniemi and TWAIL converge on a shared insight: law’s authority depends less on its content than on how it performs neutrality. As Koskenniemi’s analysis (2005, pp. 589–596) shows, legal reasoning often hides political choices beneath technical language. TWAIL adds that this habit is not innocent—it helps keep global hierarchies intact. Together, they give us a language to read today’s fascination with AI in justice as part of a longer story about law’s search for neutral authority. When policymakers describe algorithmic tools as objective or data-driven, they repeat a familiar gesture—translating political judgment into technical form. What looks like innovation may therefore be understood as a continuation of earlier patterns of depoliticisation.
The way data are collected and how categories are set up has a real effect on whose experiences end up being reflected in the system. TWAIL scholars have long pointed this out: what is described as neutral often reflects existing inequalities rather than moving beyond them. Consider an automated case-management system that prizes speed: it rewards those fluent in digital tools while pushing others to the margins (Eubanks, 2018, pp. 43–45). Predictive-analytics programs trained on old court records repeat the same pattern, reproducing bias under a new technical name. Calling such outcomes “neutral” does not make them fair—it only hides old hierarchies behind the authority of code.
This cycle mirrors Koskenniemi’s oscillation. When AI systems face moral critique—bias or opacity—reformers respond with audits, ethics charters, or impact assessments. When these are called symbolic, the defence shifts to pragmatism: efficiency and cost. The movement from moral to technical and back again reproduces the same managerial pattern that has long sustained international law (Kennedy, 2016, pp. 93–100, 106–107). Institutions appear self-correcting while leaving the structure intact—the politics of expertise under another name.
TWAIL situates this pattern within the global political economy of knowledge and production. Many judicial-AI systems are developed by private firms in the Global North, reflecting its managerial ideals of transparency and control. The result is a new form of dependency: jurisdictions in the South become users of technologies whose standards they do not shape. In this sense, what appears as universal design reproduces existing hierarchies of authority and distribution (Pahuja, 2011, pp. 365–367).
Both frameworks open a wider question about what makes authority legitimate. When expertise becomes the only ground for trust, it soon turns into power that answers to no one. Real legitimacy needs room for disagreement—moments when people can ask why the system works the way it does and who it truly serves. As Koskenniemi (2005, pp. 560–563) argues, disagreement is not law’s flaw but its condition of life. TWAIL turns this into a normative demand: participation and redistribution must replace the rhetoric of neutrality (Mutua, 2000, p. 32; Gathii, 2011, pp. 27–29). In digital justice, this suggests that systems should be open to challenge, rather than relying only on technical forms of review. Accountability concerns are addressed mainly through procedural and compliance-oriented mechanisms (Veale & Borgesius, 2021, pp. 99–105). The theoretical frame developed here thus calls for a shift from trust in systems to trust through scrutiny. AI does not escape law’s older dilemmas; it gives them a new technical form. Recognizing this continuity allows reformers to move past procedural optimism and engage technology as a political question. Addressing the histories of inequality that sit behind claims of neutrality is likely to be important if law is to sustain its legitimacy in the digital age.
When ISDS began to take shape in the 1950s and 1960s, many portrayed it as law’s victory over politics — a way to replace diplomatic bargaining with legal reasoning. As investment disputes moved from diplomacy into arbitration, this shift was described as progress toward neutral, rules-based judgment—one supposedly untouched by ideology. Supporters believed arbitration could deliver decisions grounded in expertise rather than influence — a forum where knowledge, not power, would decide.
Neutral expertise became the system’s moral vocabulary, with efficiency, predictability, and legal certainty treated as guarantees of fairness (Van Harten, 2007, pp. 28–31; Miles, 2013, pp. 112–118). In this framing, law’s legitimacy was equated with its distance from politics, and technical form became a surrogate for justice.
ISDS institutionalised this aspiration. Panels composed mostly of commercial lawyers were assumed to deliver neutral judgments by applying treaty standards such as fair and equitable treatment or legitimate expectations. The World Bank’s International Centre for Settlement of Investment Disputes (ICSID) and a growing network of bilateral investment treaties were framed as neutral instruments of global economic governance (Sornarajah, 2015, pp. 78–84). Their language was managerial rather than political—investor confidence, stability, transparency. Yet what disappeared behind that vocabulary was the asymmetry of the system itself: only investors could bring claims, and states could only defend. The professional idiom of expertise gave the appearance of even-handedness, turning structural bias into procedural order.
Critical scholarship has since shown that this self-image of neutrality was less an empirical fact than a performance sustained by institutional design and discourse. As Koskenniemi (2005, pp. 560–563) argued in another context, the claim to neutrality often masks an underlying politics of management. In ISDS, legal reasoning became a tool of economic governance, translating distributive questions into legal formulae and moral tensions into technical interpretation. To challenge that system was to question a specific mode of legitimacy—one that stabilized authority not through consent or equality but through the appearance of expertise. The early promise of neutral adjudication thus reveals itself, in retrospect, as a defining gesture of managerial legality: law’s claim to autonomy, sustained by absorbing conflict into procedural form.
Beneath its talk of neutrality, the practice of ISDS worked in one direction. Investors could initiate claims; states could only answer them. The design created what looked like parity but in practice locked public authority into a defensive posture. On paper, this was meant to balance the protection of capital with the sovereign right to regulate. In reality, it produced a very different result: it turned public-policy choices into private legal claims. As Sornarajah (2015, pp. 112–118) notes, the supposed equality between investors and states was always a myth of neutrality. What looked like procedural parity concealed a material gap—favouring those with the means to litigate, hire experts, and translate political harm into legal language.
This comes through quite clearly in the case law. In CMS v. Argentina (2005), the tribunal held Argentina liable for failing to meet investor expectations, even though the measures were taken during a serious economic crisis. In Tecmed v. Mexico (2003), a dispute that arose in an environmental context was treated as an issue of expropriation, with the tribunal focusing on the impact on the investor. The two Vattenfall v. Germany proceedings went further, turning energy and environmental policy into grounds for corporate compensation. Together, these decisions illustrated what TWAIL scholars call the structural bias of international economic law: a regime that treats investor security as universal while confining state responsibility to the local level (Titi, 2014, pp. 146–152; Anghie, 2005, pp. 245–251). The outcome was not only distributive; it reshaped how legitimacy itself was understood. States were encouraged to appear modern and predictable—to manage rather than contest the imbalance of power.
Reform efforts did little to change that script. Under UNCITRAL Working Group III and in the debates over a Multilateral Investment Court (MIC), reformers adjusted procedures while leaving the architecture intact. New transparency provisions and appellate mechanisms were introduced as improvements, yet they did not significantly alter the underlying asymmetry. More broadly, this points to a recurring tendency to address legitimacy concerns through procedural adjustments, rather than by engaging with the structural inequalities of the system (Kennedy, 2016, pp. 106–107; Sornarajah, 2021, p. 48).
From a critical-legal perspective, this asymmetry is not a flaw but a condition of the system’s legitimacy. The performance of neutrality—through procedure, expertise, and precedent—did not eliminate conflict; it domesticated it. ISDS thus perfected a form of managerial legality where disputes survive only as things to be managed. That remains its most enduring legacy—and the warning that digital justice reform should take most seriously.
If the history of ISDS shows how central expertise has been to the system, its reform trajectory suggests a more subtle pattern: criticism is rarely rejected, but often absorbed and reshaped. Concerns about bias, inconsistency, and opacity have been widely acknowledged, yet they have not led to a sustained engagement with the system’s distributive foundations. Instead, reform has mainly focused on how the system works in practice. This is most visible in efforts to introduce arbitrator codes of conduct, expand transparency, and refine procedural frameworks, particularly in discussions within UNCITRAL Working Group III (Atanasova et al., 2024, p. 1).
These initiatives tend to prioritize issues such as consistency, correctness, and the calculation of damages. At the same time, more fundamental questions about structural asymmetries remain in the background (Atanasova et al., 2024, p. 1; Sornarajah, 2021, p. 48). The effect is a gradual shift in how the problem itself is understood: what might appear as a question of inequality is recast as a matter of institutional design.
This shift is not unique to ISDS. As Kennedy (2016, pp. 2–4) observes, political disagreement is often translated into the language of technical reasoning and expert management. Once the issue is framed in these terms, legitimacy is judged more by procedural improvements than by substantive outcomes. Reform then takes the form of ongoing adjustments. It may signal progress, but it does not necessarily address the distributional structure that underlies these concerns.
Over time, this managerial logic changed how legitimacy was understood. Reform was increasingly seen as a sign of self-correction, with fairness tied to improved procedure. The dominant vocabulary—efficiency, consistency, accountability—began to resemble that of corporate governance more than public law. As Koskenniemi (2005, pp. 600–605) notes, legal arrangements may stabilise existing distributions of power, often “end[ing] up always on the side of the status quo” (p. 605).
A similar tendency is visible in current debates on digital justice. Recent European and comparative policy documents present AI tools in courts as ways to improve efficiency, accessibility, and transparency. They also link these tools to broader objectives such as legal certainty and, in some contexts, predictability (CEPEJ-AIAB, 2025, pp. 3–4; CEPEJ, 2025, pp. 3, 13–15; OECD, 2025, pp. 275–277; Cascone, 2025, p. 4). When concerns arise about bias or opacity, the proposed response tends to remain managerial: impact assessments, documentation requirements, transparency obligations, human oversight, traceability, and technical audits (CEPEJ, 2025, pp. 3, 14–15; OECD, 2025, pp. 275–277; Cascone, 2025, pp. 1–2, 7). The emphasis, in other words, shifts from the social consequences of automation to the conditions under which it may be safely administered.
In practice, opacity often remains. Making systems more visible does not necessarily make them easier to understand. AI systems often operate through multiple layers that are difficult to follow in practice (Crawford, 2021, pp. 20–22). What initially appears as a project of widening access to justice risks becoming a framework of procedural supervision. Taken together, these examples show that reform discourse consistently redirects normative concerns into questions of technical administration. This supports the broader claim advanced in this article: legitimacy is often pursued through managerial procedures rather than through more substantive shifts in authority or distribution.
The resemblance to ISDS is therefore not merely rhetorical. In both settings, appeals to neutrality leave basic questions unresolved: who benefits, how authority is distributed, and whose interests remain privileged. ISDS sought legitimacy through arbitral professionalism; digital justice does so through design ethics, auditability, and procedural safeguards. In both cases, political disagreement is translated into a technical problem to be managed by experts. What recedes from view is not procedure itself, but the distributive choices embedded in the system.
Recognising this continuity does not require rejecting technology. It requires treating the rise of AI in adjudication as a moment that reshapes the terms of legal authority. The comparison between ISDS and digital justice is not about institutional similarity, but about a shared way of responding to critique. In both settings, legitimacy crises are addressed not by redistributing authority, but by reorganising how decisions are produced and justified. In ISDS, this took the form of procedural reform—transparency rules, appellate mechanisms, and codes of conduct. In digital justice, it appears through auditability, risk classification, human oversight, and technical standards. What links the two is a common logic: political disagreement is translated into questions of design and administration. It is this continuity of managerial response that makes the comparison analytically meaningful—and that raises a further question: whether justice can remain accountable, contestable, and publicly intelligible when legitimacy is increasingly measured through managerial control.
What passes for reform in international law rarely breaks with the past; it usually rearranges it. As Koskenniemi (2005, p. 605) suggests, because legal meaning depends on practice rather than inherent structure, legal arrangements do not necessarily transform underlying relations of power and may instead stabilize them. Reform is not the opposite of continuity but one of the ways it is maintained.
The ISDS experience shows this quite clearly. As criticism of arbitral practice intensified—especially around bias, inconsistency, and transparency—the response was not to question the system itself. It was to modify how it operates. The European Union’s Investment Court System (ICS), and later the Multilateral Investment Court (MIC) proposal, were introduced in this vein: standing adjudicators, an appellate mechanism, and additional procedural safeguards (Titi, 2022, pp. 84–88).
But these changes do not really take the system outside its original logic. They reorganize it. The underlying commitment to investment protection — and to a depoliticized, rule-based framework — remains intact. In that sense, what is presented as reform looks more like adjustment. This becomes even clearer when one considers that the IIA regime is still largely shaped by old-generation treaties, which continue to define the basic terms of the relationship between investors and host states (UNCTAD, 2023, p. 71).
This pattern reveals a deeper grammar of legitimacy. The call for reform becomes a way of staging responsiveness while avoiding redistribution. Technical improvement substitutes for political debate. Once more, transparency and efficiency take the place of justice itself. In this sense, reform often functions as a way of signalling responsiveness: it suggests that international law can learn, adapt, and self-correct. Yet the object of correction is almost always procedural form, not structural hierarchy. The process becomes the proof.
The reform projects that followed the ISDS crisis did not replace managerial legality; if anything, they reinforced it. The European Union’s ICS and the proposed MIC were introduced as correctives to arbitral practice. But they did not move the system outside its existing logic. What shifted was not the distribution of authority, but the way the system justified itself.
This reflects a broader tendency identified by Kennedy: problems of global governance are often approached as matters of technical management and expert intervention (Kennedy, 2016, pp. 106–107). Reform, in this setting, tends to take the form of institutional design rather than political choice. Independence, transparency, predictability—these are treated as obvious goods. But they also narrow the space of disagreement by relocating it into procedure. Once the process appears orderly, the harder questions tend to recede.
From a critical perspective, this does not resolve conflict so much as displace it. Koskenniemi’s account of indeterminacy is useful here. Legal reasoning does not eliminate underlying tensions; it works through them. Any resolution depends on choices that remain open to contestation (Koskenniemi, 2005, pp. 589–596). Reform, then, stays within those limits.A similar point appears in TWAIL scholarship. As Pahuja shows, claims to universality often present particular arrangements as general and necessary, extending their reach while masking their uneven effects (Pahuja, 2011, pp. 365–367). Seen from this angle, the move from ISDS to ICS or MIC does not amount to a break. The structure largely remains; what changes is how it is described and defended.
The language of reform has long been tied to promises of innovation. In recent years, that promise has shifted from investment law to debates on civil justice and digital governance. The vocabulary has changed—AI, automation, data-driven efficiency—but the underlying logic is familiar. Much of the current policy and academic discussion presents AI in fairly practical terms: as a way to make systems run faster, handle growing caseloads, and improve administrative performance (OECD, 2025, p. 5; Guitton et al., 2025, p. 1; Skutele, 2025, p. 167). It is often introduced as a response to pressure on institutions rather than as a reason to question how they are structured. In that sense, legitimacy becomes tied to improvement, not transformation.
The appeal of innovation rests on an older assumption—that better design can fix what governance struggles to handle. Terms such as transparency, accountability, or human-centred AI appear again and again in policy texts. They are rarely contested; they tend to be taken for granted. But this also shifts the discussion. Questions that might once have been argued in political terms start to look like matters of implementation and technical adjustment.
The same pattern is visible in current reform agendas. Discussions on digital courts and AI-assisted adjudication tend to revolve around modernisation, with efficiency, case management, and procedural optimisation coming to the fore (CEPEJ, 2025, p. 3). This does not remove disagreement so much as change the way it appears. Legal reasoning still has to deal with underlying tensions. As Koskenniemi shows, those tensions are not resolved but worked through, often in a more technical language that makes the choices involved seem less visible (Koskenniemi, 2005, pp. 589–596). Innovation, in this sense, does not simply transform the system. It reshapes the way it is justified.
This pattern becomes clearer when policy discourse is read alongside institutional practice. Recent CEPEJ findings identify over one hundred AI and cyberjustice tools already in use across European judicial systems, primarily designed to improve efficiency and accessibility (CEPEJ, 2025, pp. 3–4). These tools support functions such as document processing, case management, information services, and decision support, with increasing reliance on machine learning and natural language processing (CEPEJ, 2025, pp. 13–15). At the same time, guidance continues to frame their use in terms of operational gains while emphasising safeguards such as human oversight, transparency, and traceability (CEPEJ, 2025, pp. 3, 13–14).
A similar dynamic appears beyond the judicial field. OECD (2025, pp. 275–277) shows that public-sector AI deployments are primarily oriented toward administrative optimisation tasks, such as workflow automation and decision support, while concerns about transparency, trust, and accountability persist. The regulatory response often follows the same logic. The EU AI Act, for instance, structures governance around risk classification, documentation, and compliance, aiming to make AI “trustworthy,” while leaving existing distributions of authority largely intact (European Commission, 2021, pp. 2–3, 12–14; Veale & Borgesius, 2021, pp. 99–102).
Taken together, these developments suggest that digital justice reform, much like earlier ISDS reform, addresses legitimacy concerns primarily through the refinement of process. What appears as technical improvement is also a redistribution of authority into design choices, metrics, and institutional infrastructures, rather than a transformation of underlying structures.
In some contexts, these dynamics become more visible at the level of individual experience. Studies of automated welfare and administrative systems show how efficiency-oriented tools can produce outcomes that are difficult to contest, particularly for those already in vulnerable positions (Eubanks, 2018, pp. 43–47). What appears as neutral categorization can translate into concrete disadvantages when individuals are unable to understand or challenge how decisions are made. This illustrates how technical systems do not simply improve administration; they also reshape how authority is experienced in practice. Such dynamics have already been observed in systems such as automated welfare allocation and risk-scoring tools, where efficiency gains often come at the cost of contestability and transparency (Yeung, 2018, pp. 507–510).
For Koskenniemi (2005, pp. 589–596), this dynamic reflects law’s constant movement between utopia and apology: moments of critique invite moral language—fairness, transparency, participation—only to be stabilised through technical forms of governance. Kennedy (2016, pp. 12–14) similarly describes a managerial politics of expertise, in which disagreement is reframed as a problem of design and administration.
TWAIL scholarship grounds this pattern in global inequality. As Pahuja (2011, pp. 365–367) shows, claims to universality often extend particular arrangements while at the same time masking their uneven effects. From this perspective, digital justice reforms, like earlier investment-law reforms, risk reproducing existing hierarchies—modernising institutional form while leaving underlying distributions of authority largely intact.
Modern legal systems live with a built-in tension. They must compel obedience, yet also persuade that such power is justified. As Koskenniemi (2005, pp. 560–563) notes, legality shifts constantly between two registers—technical authority and moral aspiration. When power begins to look like domination, law borrows the language of justice; when justice provokes disagreement, it retreats to the safety of rules. Authority often depends on being perceived as legitimate, and this legitimacy is frequently maintained through procedures that recast political conflict in technical terms.
This tension tends to surface most clearly at moments of reform. As Kennedy (2016, pp. 1–3) notes, expertise does not resolve conflict so much as change how it is articulated and handled. In both investment arbitration and digital justice, reform efforts often recast broader normative concerns in procedural terms. Oversight mechanisms, ethics codes, and transparency tools can give the impression that authority regulates itself, even as they help sustain it in practice.
A TWAIL perspective adds an important spatial dimension to this picture. What counts as legitimate authority does not emerge in a vacuum but reflects the material and epistemic hierarchies of the international order. As Anghie (2005, pp. 245–251) and Pahuja (2011, pp. 365–367) show, claims to universality can obscure these hierarchies, presenting particular governance practices as generally applicable. Seen in this context, reform projects are closely tied to the settings from which they emerge. When improvement is defined through benchmarks such as efficiency, predictability, and managerial control, there is a risk that existing asymmetries are carried forward rather than addressed.
AI-driven reform illustrates this pattern with particular clarity. Automated systems are often presented as neutral because they reduce human discretion. In practice, they shift interpretive authority away from judges and toward designers, engineers, and private actors. Judges and litigants become users of systems they cannot fully contest. What appears as technological progress becomes, in Koskenniemi’s terms (2005, pp. 589–596), a familiar apologetic move—reaffirming existing power through the language of innovation.
Seen in this light, digital reform does not resolve the tension between authority and legitimacy; it reorganises how that tension is managed.
If legitimacy is the story law tells about itself, authority refers to the practices through which that claim is sustained. The distinction matters because, in periods of reform, legitimacy is often framed in normative terms, while authority is reorganized at the level of technique.
In the digital age, much of this reorganization unfolds through technical systems. AI in civil justice is often introduced as a way to reduce bias and improve consistency. Whether such claims hold empirically is less important here than the role they play within reform discourse: they present technical intervention as a response to concerns about fairness.
As Cohen (2019, pp. 232–234) shows, when governance becomes embedded in technical infrastructures, decision-making processes tend to grow less accessible to external scrutiny. The capacity to question outcomes is correspondingly reduced. Authority does not disappear; it shifts location—from courtroom reasoning to system design. What matters is no longer only how decisions are made, but how systems are constructed and what they allow or exclude.
This shift is also visible in how expertise is distributed. Authority increasingly moves toward programmers, data scientists, and private vendors. Benjamin (2019, pp. 89–92) describes this as the “New Jim Code,” where inequalities are reproduced through technological design rather than formal rules.
Empirical studies illustrate how these dynamics unfold in practice. Eubanks (2018, pp. 43–47) documents how automated welfare systems categorize vulnerable populations in ways that are difficult to contest. What appears as neutral classification can translate into concrete disadvantage. In similar ways, court technologies that prioritize efficiency—such as automated triage or predictive tools—may narrow the space for interpretation and challenge.
From a broader perspective, Zuboff (2019, pp. 111–115) describes how data-driven systems extend forms of control through prediction and behavioural steering. Comparable dynamics are now visible in digital justice, where private infrastructures shape how authority is exercised. As Veale and Borgesius (2021, pp. 99–105) note, transparency alone does not ensure accountability; systems may be partially visible yet remain difficult to challenge in practice.
Together, these developments point to a shift in how authority operates: increasingly embedded in technical systems, less visible, and harder to contest.
AI extends law’s longstanding reliance on technical expertise, but does so in ways that further obscure how decisions are shaped. As Kennedy (2016, pp. 6–8) suggests, governance often recasts political disagreement as technical problem-solving. The digital turn intensifies this tendency. Algorithms do not simply apply legal reasoning; they also shape what counts as a reasonable decision.
Koskenniemi’s (2005, pp. 560–563) work helps situate this shift. Legal argument has long moved between normative aspiration and technical form, and moments of contestation tend to push institutions toward the latter. When legitimacy is questioned, institutions respond by adding procedures—audits, guidelines, reporting frameworks. AI fits easily into this pattern, presenting fairness as something that can be secured through calculation, while the assumptions embedded in the system become less visible.
From a TWAIL perspective, this shift is easier to see in its global context. As Anghie (2005, pp. 245–251) and Pahuja (2011, pp. 365–367) have shown, what is presented as universal often reflects particular configurations of power, even when it appears neutral. Many of the tools now shaping legal decision-making—programming languages, data models, and measures of transparency—are developed in the Global North and carry with them certain assumptions about order and rationality. When these systems are adopted elsewhere, they do not simply transfer technology; they also carry those underlying assumptions, which can, in practice, reinforce existing dependencies.
Seen this way, the limits of transparency become clearer. As Crawford (2021, pp. 20–22) notes, visibility is often mistaken for understanding. Complex systems may reveal parts of their operation while obscuring their overall logic. The criteria that shape outcomes—what is counted, weighted, or ignored—remain difficult to access or interpret. As more decisions are mediated through such systems, expertise becomes both the medium and, at times, a shield for power.
Algorithmic legality does not remove politics; it reshapes how it operates through technical systems. The question is not only how such systems operate, but whether those affected can meaningfully challenge them.
Across critical scholarship, a common concern is that reform may fall short of justice if the structures sustaining inequality are left intact. Koskenniemi (2005, pp. 589–596) points out that law’s legitimacy often rests on its capacity to render conflict into forms that appear ordered. Yet when the grammar of order itself sustains exclusion, refinement cannot deliver renewal. Post-Koskenniemi scholars argue, following Orford (2011), Pahuja (2011), and Johns (2013), that reform becomes meaningful only when it breaks with law’s technocratic habits— habits that Koskenniemi (2005, pp. 589–596) associates with the stabilizing pull of technical reasoning.
Across these critical traditions, the call is similar: reform must open itself to plurality and to the voices it has long disciplined. Marks (2000, pp. 24–26) suggests that treating reform as a technical exercise misses something more fundamental—how authority is organised in the first place. Krisch (2010, pp. 31–34) picks up on this by describing a pluralist order in which authority is not concentrated in a single centre but spread across different sites. Rajagopal (2003, pp. 79–83) brings this perspective into a postcolonial setting, showing how reform efforts can end up reproducing the very hierarchies they aim to overcome.
TWAIL extends this critique by restoring history and geography to the debate. Building on Rajagopal (2003, pp. 79–83) and Pahuja (2011, pp. 365–367), TWAIL scholars and critical international lawyers show how claims to global reform often reproduce the hierarchies they seek to address. Digital justice initiatives, like investment-law reforms before them, carry traces of this universalism, privileging particular forms of expertise while presenting them as neutral. A genuine alternative would start from those most affected by these arrangements—communities in the Global South and others positioned at the receiving end of legal and technological interventions, where law’s effects are felt most sharply and its authorship least visible. If reform is to matter, it must remain open to such perspectives, treating critique not only as correction but as a source of alternative legal imaginaries.
Taken together, these perspectives shift the focus from efficiency to accountability, from performance to participation. Kennedy (2016, pp. 4–7) shows how expertise speaks in the name of neutrality and universal reason, even as it helps translate political conflict into technical disagreement. For both TWAIL and post-Koskenniemi thinkers, critique is not an act of refusal but a form of creation—it keeps law open, unfinished, and answerable. Feminist critiques add another dimension by showing how law’s universal language can obscure gendered and embodied experiences of inequality. As Charlesworth (1999, pp. 381–382) argues, attending to plurality in international law also requires exposing the silences that sustain it.
For reform to matter, it must take this openness seriously. The next section turns from critique to construction: how civil justice reform might reclaim legitimacy through rights, participation, and democratic oversight—what this paper calls a human-centred justice.
If legitimacy lives through practice, then justice itself must remain recognizably human. Automation speaks the language of confidence—promising reliability, speed, and accuracy as substitutes for trust. Yet people trust law not because it is flawless but because it listens, explains, and recognizes them. Fuller (1969, pp. 39–42) suggests that legality carries moral force only when rules remain intelligible to those subject to them. An algorithm may follow rules perfectly and still fail this test: such a system obeys without understanding.
Human-centred justice turns that logic around. It asks how institutions can keep space for judgment, story, and empathy inside digital infrastructures. Cappelletti and Garth (1978, pp. 203–208) emphasize that access to justice involves meaningful participation, not merely formal access to courts. When design takes the place of dialogue, legitimacy shrinks into procedure. What gives justice its meaning is not seamless process but the act of being heard.
Europe’s new policy experiments capture both the promise and the limits of this turn. The CEPEJ Ethical Charter and recent OECD work on AI governance emphasize principles such as user-centred design, transparency, and trustworthy automation (CEPEJ, 2018; OECD, 2025). As Floridi and Cowls (2019, pp. 690–694) propose in their unified framework for AI ethics, trustworthiness depends not only on technical reliability but also on principles of justice, accountability, and human oversight. Crawford (2021, pp. 134–136) shows that focusing on technical fixes can obscure deeper structural questions about how AI systems reproduce inequality. Principles are adopted, reports are published—yet few spaces exist where those affected can shape decisions. Dialogue becomes consultation; accountability turns into documentation. Trust built on paperwork seldom endures.
The European Commission’s Proposal for an Artificial Intelligence Act (COM/2021/206) and the CEPEJ (2018) Ethical Charter share the same aspiration: to embed accountability into the architecture of automation while preserving the legitimacy of judicial authority.
TWAIL scholarship widens the view. Technologies, like laws, are never neutral; they inherit the geography and bias of their makers. Santos (2014, pp. 16–18) calls this the struggle for epistemic justice—the right to define what counts as knowledge. A genuinely human-centred digital justice must open design itself to plural voices. Otherwise, digitalization will repeat the hierarchies it claims to correct.
Such justice does not reject technology; it reclaims it as a medium of encounter. Explainability becomes more than a software function—it becomes a civic right: the right to know how one is judged and to question it. In that sense, legitimacy moves from system trust to human trust—from belief in code to belief in conversation.
This brief analogy situates the argument within a broader governance context. It does not extend the ISDS–AI comparison developed in this article, but illustrates how similar managerial logics appear beyond the field of judicial reform.
Recent debates on climate justice offer a useful parallel. Much like digital justice initiatives, climate governance is often framed through technical solutions—efficiency improvements, carbon metrics, and data-driven monitoring—presented as neutral responses to complex global problems. In practice, attention often shifts from distribution to management, with a focus on measurement and optimisation while underlying inequalities persist (Crawford, 2021, pp. 8–12).
This dynamic becomes visible in the infrastructures that sustain digital systems. The environmental costs of data infrastructures—from energy-intensive data centres to the extraction of rare-earth minerals—are often treated as external technical concerns rather than as questions of justice. Similarly, in both climate governance and digital justice, evidence increasingly takes technical form: datasets, indicators, and predictive models shape what counts as harm and how it is recognized. As Crawford (2021, pp. 20–22) shows, such systems do not merely represent reality; they structure how it is perceived and acted upon.
What links these domains is not their subject matter, but a shared tendency to manage legitimacy through technical means. In both cases, contestability becomes limited not by the absence of rules, but by the way problems are framed. When decisions are mediated through systems that are difficult to interrogate in practice, they often come to be seen as objective (Veale & Borgesius, 2021, pp. 99–105).
The point of this analogy is not to expand the scope of the article, but to underline a recurring pattern: across domains, legitimacy is often pursued through technical responses, while underlying inequalities tend to persist.
This article has argued that the rise of digital justice does not mark a break from earlier forms of legal governance. Rather, it extends a familiar pattern in which legitimacy crises are addressed through managerial reform. The comparison with ISDS shows how such reforms operate: not by redistributing authority, but by reorganising how it is exercised and justified.
In digital justice, this shift shows up in fairly practical ways. AI tools are often introduced to make processes faster and more consistent, and to reduce pressure on courts. At the same time, parts of legal decision-making shift into less visible domains, such as system design, data selection, and technical infrastructure. Improvements at this level can make systems work but may also carry forward existing imbalances.
This is why the question is not only how AI should be regulated. It also concerns how authority is exercised when decisions increasingly rely on technical systems. Measures such as transparency requirements, audits, and human oversight matter, but on their own they do not settle the question. Much depends on whether people can understand how decisions are made and, where necessary, question them.
This concern extends beyond design to how these systems are defined in practice and who has a say in that process. It raises fairly direct questions: what criteria are used, whether they can be challenged, and whose perspectives are actually taken into account. Dealing with this means putting in place arrangements that allow decisions to be examined and, where necessary, questioned and revised.
As AI becomes more embedded in legal processes, questions of legitimacy are likely to turn on these practical conditions: not simply on whether systems are described as neutral, but on whether those affected can still engage with and challenge the outcomes. In that sense, the direction of digital justice will depend on how far legal systems remain open to scrutiny and change, rather than on how far they can be optimised. What is at stake is not only the efficiency of legal systems, but the conditions under which authority can remain accountable and open to contestation. This is not offered as a guaranteed solution, but as a minimal condition for avoiding the repetition of the same managerial patterns of legitimacy.
Anghie, A. (2005). Imperialism, sovereignty and the making of international law. Cambridge University Press.
Atanasova, D., Ostřanský, J., & Beyer, V. (2024). Compensation and damages in investor-state dispute settlement: Options for reform. International Institute for Sustainable Development. https://www.iisd.org/publications/report/compensation-damages-isds-reform
Benjamin, R. (2019). Race after technology: Abolitionist tools for the new Jim Code. Polity Press.
Cappelletti, M., & Garth, B. (1978). Access to justice: The worldwide movement to make rights effective. Sijthoff and Noordhoff.
Cascone, G. (2025). The use of artificial intelligence in EU criminal justice systems: First insights and emerging trends. Open Research Europe. https://open-research-europe.ec.europa.eu/articles/5-361
Charlesworth, H. (1999). Feminist methods in international law. American Journal of International Law, 93(2), 379–394.
Cohen, J. E. (2019). Between truth and power: The legal constructions of informational capitalism. Oxford University Press.
Council of Europe, CEPEJ. (2018). European ethical charter on the use of artificial intelligence in judicial systems and their environment. https://rm.coe.int/ethical-charter-en-for-publication-4-december-2018/16808f699c
Council of Europe, CEPEJ. (2025). 1st report on the use of artificial intelligence (AI) in the judiciary. https://rm.coe.int/cepej-aiab-2024-4rev5-en-first-aiab-report-2788-0938-9324-v-1/1680b49def
Council of Europe, CEPEJ-AIAB. (2025). Reflections on the use of artificial intelligence in judicial systems. https://rm.coe.int/cepej-aiab-2025-1rev5-en-reflections-of-the-aiab-on-the-use-of-artific/1680b42a56
Crawford, K. (2021). Atlas of AI: Power, politics, and the planetary costs of artificial intelligence. Yale University Press.
Eubanks, V. (2018). Automating inequality: How high-tech tools profile, police, and punish the poor. St. Martin’s Press.
European Commission. (2021). Proposal for a regulation laying down harmonised rules on artificial intelligence (Artificial Intelligence Act) (COM/2021/206 final). https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:52021PC0206
Floridi, L., & Cowls, J. (2019). A unified framework of five principles for AI in society. Nature Machine Intelligence, 1(9), 689–699. https://doi.org/10.1038/s42256-019-0088-2
Fuller, L. L. (1969). The morality of law (Rev. ed.). Yale University Press.
Gathii, J. T. (2011). Third World approaches to international economic governance. In D. Armstrong (Ed.), Routledge handbook of international law (pp. 27–45). Routledge.
Guitton, M., Druta, V., Hinterleitner, M., Tamò-Larrieux, A., Landuyt, D., Fosch-Villaronga, E., Kamara, I., & Pałka, P. (2025). Adoption of artificial intelligence in the judiciary: A comparison of 28 advanced democracies. Discover Artificial Intelligence, 5, Article 169. https://doi.org/10.1007/s44163-025-00311-y
Johns, F. (2013). Non-legality in international law: Unruly law. Cambridge University Press.
Kennedy, D. (2016). A world of struggle: How power, law, and expertise shape global political economy. Princeton University Press.
Koskenniemi, M. (2005). From apology to utopia: The structure of international legal argument. Cambridge University Press.
Krisch, N. (2010). Beyond constitutionalism: The pluralist structure of postnational law. Oxford University Press.
Marks, S. (2000). The riddle of all constitutions: International law, democracy, and the critique of ideology. Oxford University Press.
Miles, K. (2013). The origins of international investment law: Empire, environment and the safeguarding of capital. Cambridge University Press.
Mutua, M. (2000). What is TWAIL? Proceedings of the ASIL Annual Meeting, 94, 31–38. https://doi.org/10.1017/S027250370005239X
OECD. (2025). Governing with artificial intelligence. OECD Publishing. https://www.oecd.org/content/dam/oecd/en/publications/reports/2025/06/governing-with-artificial-intelligence_398fa287/795de142-en.pdf
Orford, A. (2011). International authority and the responsibility to protect. Cambridge University Press.
Pahuja, S. (2011). Decolonising international law: Development, economic growth and the politics of universality. Cambridge University Press.
Rajagopal, B. (2003). International law from below: Development, social movements, and Third World resistance. Cambridge University Press.
Santos, B. de S. (2014). Epistemologies of the South: Justice against epistemicide. Paradigm Publishers.
Skutele, S. (2025). Using artificial intelligence to enhance judicial efficiency from the perspective of the right to a fair trial. Teisė, 137, 167–179. https://doi.org/10.15388/Teise.2025.137.11
Sornarajah, M. (2015). Resistance and change in the international law on foreign investment. Cambridge University Press.
Sornarajah, M. (2021). The international law on foreign investment (5th ed.). Cambridge University Press.
Titi, C. (2014). The right to regulate in international investment law. Nomos/Hart Publishing.
Titi, C. (2022). The European Union’s investment court system. Oxford University Press.
UNCTAD. (2023). World investment report 2023. United Nations. https://unctad.org/publication/world-investment-report-2023
Van Harten, G. (2007). Investment treaty arbitration and public law. Oxford University Press.
Veale, M., & Borgesius, F. Z. (2021). Demystifying the draft EU Artificial Intelligence Act. Computer Law Review International, 22(4), 97–112. https://doi.org/10.9785/cri-2021-220402
Yeung, K. (2018). Algorithmic regulation: A critical interrogation. Regulation & Governance, 12(4), 505–523. https://doi.org/10.1111/rego.12158
Zuboff, S. (2019). The age of surveillance capitalism: The fight for a human future at the new frontier of power. Profile Books.
CASES
CMS Gas Transmission Company v. The Argentine Republic, ICSID Case No. ARB/01/8, International Centre for Settlement of Investment Disputes, May 12, 2005.https://www.italaw.com/cases/288
Técnicas Medioambientales Tecmed, S.A. v. The United Mexican States, ICSID Case No. ARB(AF)/00/2, International Centre for Settlement of Investment Disputes, May 29, 2003. https://www.italaw.com/cases/1087
Vattenfall AB, Vattenfall Europe AG, Vattenfall Europe Generation AG v. Federal Republic of Germany, ICSID Case No. ARB/09/6, International Centre for Settlement of Investment Disputes, March 11, 2011. https://www.italaw.com/cases/1148
Vattenfall AB and others v. Federal Republic of Germany, ICSID Case No. ARB/12/12, International Centre for Settlement of Investment Disputes. https://www.italaw.com/cases/1654
* ORCID: 0000-0002-3898-0505